Classification of cancer cells and gene selection based on microarray data using MOPSO algorithm

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray data analysis is vital for cancer research and classification.
• Limited sample sizes in microarrays pose challenges for accurate cancer classification.
• Effective gene selection is crucial for improving cancer classification accuracy.

Purpose of the Study:

• To develop a novel hybrid model for classifying high-dimensional microarray data.
• To enhance cancer classification accuracy using an optimized gene selection approach.
• To address the challenge of limited sample sizes in cancer microarray datasets.

Main Methods:

• A hybrid model combining multi-objective particle swarm optimisation (MOPSO) with linear Bayesian discriminant analysis.
• Utilizing a binary vector representation where each bit corresponds to a gene's selection status.
• Assessing gene set quality using a classification algorithm to determine particle fitness.

Main Results:

• The proposed MOPSO algorithm was applied to four distinct cancer databases.
• Achieved an average improvement of 25.84% in classification accuracy across the databases.
• Demonstrated significant accuracy improvements in blood (18.63%), lung (24.25%), breast (27.73%), and prostate (32.80%) cancer datasets.

Conclusions:

• The MOPSO-based approach effectively improves cancer data classification accuracy.
• The model reduces redundancy by identifying and selecting informative gene subsets.
• This method enhances classification by considering gene correlations and eliminating irrelevant genes.